Despite their belief to the contrary the executive branch is in charge of very little in this country. They are harassing and extorting in legally dubious and often outright illegal ways, but companies and institutions and individuals are getting wise to the fact there’s very little power these guys really have because the law is structured to prevent executive abuse of power. All you have to do is get your suit filed and get a stay, and sooner or later the governments case likely falls apart. It’s frictionful for everyone involved and will sooner or later cause serious damage to the economy, but increasing as the initial shock fades, everyone is realizing the president is fairly weak and his antics and his hand picked but of loonies undermine any power they might have. It’s not Howard Lutnicks world, and as time goes on it becomes less and less so as they squander the reputation of the presidency tilting at windmills.
This is self serving nonsense. “ai infrastructure” here isn’t manufacturing its enormous data centers that have so few humans the lights are off in every area other than where physical maintenance is happening. It’s about building enormous caverns that consume enormous amounts of power and warehouse the products his company manufactures offshore. The only thing he wants to onshore is spending on those offshore made products he sells. There’s no meaningful jobs in “AI infra” that outlive for the original data center build. He’s advocating for huge investment in energy production to ensure there’s capacity to run his products he’s selling that his company produces off shore.
The figure of speech is about a typical divers lifetime. I believe you’re confusing it with “once a generation,” which refers to the collective human experience type of rarity.
I’ll be honest, I’ve never heard that phrase in that context. My only real frame of reference is the 1981 Talking Heads hit single which I always took to have the “once in a generation” meaning. What’s a diver’s lifetime?
To be fair stock price is not a great indicator of much. Sales are down YoY with no prospect of recovery. Stock price has no bearing on current operations unless they raise capital by selling equity. It’s also been the case that Tesla stock price has been disconnected from performance often. The amount of ill will he’s generated among his core market is enormous, and the tailwinds he had with green policy have been replace with significant head winds with an aggressively (and bizarrely) pro-pollution administration based on a retribution model than a technocratic model. This is especially bad for Tesla as musk actively sought retribution so there is no reason to believe this is going to go well for Tesla over the next few years. He dumped all over his customers to curry favor with someone that now hates him and has a history of actively seeking revenge - what a mess.
As an owner of two Tesla and a frequent user of autopilot with FSD, I don’t believe we will be seeing a significant rollout in any but the easiest to travel areas. SF, LA, Seattle or the other places Waymo currently is either running or preparing to operate are much more challenging road environments than Austin, and Waymo is already fully autonomous with years of track record in these areas. There’s no chance robotaxi will overtake Waymo - I use them often, use FSD often. They aren’t comparable - Waymo is considerably more advanced and capable.
Sorry man, I wished Tesla was going to win out, but by shifting focus to ever more bizarre efforts away from their core cars (when’s the last time a real X or S model refresh happened?) they lost the initiative. BYD and the other legion of Chinese EV are going to eat Tesla alive within 5 years. I’ve had the chance to ride in and drive many of the brands overseas and they are a fraction the price and about as good. It’s over.
Stock price is certainly a better indicator than any numbers coming out of China. The question isn't whether the numbers are fudged but how many times were they fudged, at each step in the reporting chain.
Yeah this stuck out at me - the hubris of the stateless web stack supersedes the 18 years of hard unsung work at building and end to end stateful pipeline that ties out to the penny and handles all the complex business logic and reconciliations seamlessly across god knows how many integrations. No fancy diagrams or pictures of the nameless faceless heroes that had accomplished that act of heroism. For sure recognizing the value is something to trumpet, but that’s the Herculean hero story I want to hear - the DOGE bros who tied it all together with JavaScript frameworks, yawn.
The only way the database could be harnessed to do something useful is after all the people who were standing in the way in management for the last 18 years likely having been sacked. You can bet any useful project to put it to use was blocked by paper-pushers threatened by the spectre of automation, until most people had forgotten about it.
Nobody believes the database sprung forth from the earth or was created accidentally. The fact that 18 years later that project had borne no visible fruit, and that most people who could have used it, didn’t even know about it, is proof of the problem. It’s a problem of terrible management. That is what, regardless of your politics, is being slightly jostled by DOGE. Personally I have dealt with enough of our absurd government processes that I don’t think they can make anything much worse, and it cannot be less efficient.
The challenge with this way of thinking is what handicaps a lot of cultures education systems - they teach how to find the answer to a question - but that’s not where the true value lies. The true value comes from learning how to ask the right question. This is becoming even more true faster and faster as AI becomes better at answering questions of various sorts and using external tools to answer what its weak at (optimizations, math, logic, etc).
You don’t learn how to ask the right questions by just having facts at your fingertips. You need to have lots of explorations of what questions can be asked and how they are approached. This is why when you explore the history of discovery humanist societies tend to dominate the most advanced discoveries. Mechanical and rote practical focus yields advances of a pragmatic sort limited to what questions have been asked to date.
Removing arts, culture, philosophy (and its cousin politics) from assistive technologies will certainly help churn out people who will know answers, but answers the machines know better. But will not produce people who will ask questions never asked before - and the easy part of answering those questions will be accelerated with these new machines that are good at answering questions. Such questions often lie at the intersection of arts, culture, philosophy, and science - which is why Liebnitz, Newton, Aristotle, et al were polyglots across many fields asking questions never yet dreamed of as a result of the synthesis across disciplines.
I see, so service to the country’s security is not nearly as important as fealty to one man. I guess this checks out with recent deportation of decorated soldiers. What a mess we’ve become, god bless the leaders who have to fix what we’ve broken.
Yeah exactly. This is where an LLM could really shine. The trick though is consistency and that it’s often more on the basis of how the organization typically treats something and rationale to its applicability to GAAP. The creation and consistent adherence to internal standards and providing them and proving them to auditors is the key and LLMs would need infra to accomplish this.
Especially since it decomposes the image into a semantic vector space rather than the actual grid of pixels. Once the image is transformed into patch embeddings all sense of pixels is entirely destroyed. The author demonstrates a profound lack of understanding for how multimodal LLMs function that a simple query of one would elucidate immediately.
The right way to handle this is not to build it grids and whatnot, which all get blown away by the embedding encoding but to instruct it to build image processing tools of its own and to mandate their use in constructing the coordinates required and computing the eccentricity of the pattern etc in code and language space. Doing it this way you can even get it to write assertive tests comparing the original layout to the final among various image processing metrics. This would assuredly work better, take far less time, be more stable on iteration, and fits neatly into how a multimodal agentic programming tool actually functions.
Yeah, this is exactly what I was thinking. LLMs don't have precise geometrical reasoning from images. Having an intuition of how the models work is actually.a defining skill in "prompt engineering"
Asking your favorite LLM actually helps a lot. They generally are well trained on LLM papers unsurprisingly. In this case though it’s important to realize the LLM is incapable of seeing or hearing or reading. Everything has to be transformed into a vector space. Images are generally cut into patches (like 16x16) which are themselves transformed by several neural networks to convert them into a semantic space represented by the models parameters.
But this isn’t hugely different than your vision. You don’t see the pixel grid either. You have to use tools to measure things. You have the ability over time to iteratively interact with the image by perhaps counting grid lines but the LLM does not - it’s a one shot inference against this highly transformed image. They’ve gotten better at complex visual tasks including types of counting, but it’s not able to examine the image in any analytical way or even in its original representation. It’s just not possible.
It can however make tools that can. It’s very good at working with PIL and other image processing libraries or even writing image processing code de novo, and then using those to ground itself. Likewise it can not do math, but it can write a calculator that can do highly complex mathematics on its behalf.
I’d note they’re not mutually exclusive revenue streams and both add meaningfully to their value. I think the reality is they peaked the first one and growth is in the second one. Subscriptions that are sticky however are much more valuable individually than an advertising tier user. But if you can cater to both and not downgrade subscriptions to ads tier you win in two parallel markets via the same platform. This is not a bad business strategy. But they need to not lose the subscriptions and their reason for being in the quest for growth or they’ll see nominal growth with decline in value.
note: I hate ads so I'm not trying to manifest this, but can you explain why you're so sure of this?
To me, it seems like they "should" (for greed reasons, I mean, not for my happiness) hike the prices of subscriptions aggressively while keeping the ad-tier attractively-priced, moving as many people as possible over. This increases ad revenue and allows more YoY growth if their ML can manipulate you into more watch hours in 2027 than you do in 2026.
Sure, some people like me will probably drop Netflix before they'll pay $35 a month or endure ads. But the current delta is only $10. I suspect they can make $10 a head in ad revenue in a year -- and if they can make $15, they would break even if they lost 3 ad-free subscribers but gained 2 back onto the ad tier. Anything better than those numbers would be a net gain.
Because the way subscription revenue is accounted for is a present value of the expected duration of the subscription and ad revenue is cyclic and varies throughout the year across individuals, cohorts, and the population. They’re also generally different markets - people willing to endure ads are either unable to afford the subscription or cheap, in either case it’s not unreasonable to expect the impression value for ads is pretty low.
Welcome to 2022 or so. I thinl Hulu started it, but yea. Many "premium" services are back to ads again on the lowest tier. It's probable more expensive than the highest tier 10 years ago as well.
reply